Structured Text Translation

4. The system of claim 2, wherein the query vector, the key vector, and the value vector are generated in parallel.

5. The system of claim 2, wherein the multi-head attention network has a variable number of heads, one head for generating a vector by applying a weight to one of the embeddings.

8. The system of claim 6, wherein the embeddings of the tokens in the structured translated text are a second query input, a second key input, and a second value input to the first multi-head attention network of the decoder.

9. The system of claim 6, wherein the output of the first multi-head attention network corresponds to a third query input received by the second multi-head attention network, and the encodings from the encoder correspond to a third key input and a third value input received by the second multi-head attention network.

13. The method of claim 11, wherein the query vector, the key vector, and the value vector are generated in parallel.

14. The method of claim 11, wherein the multi-head attention network has a variable number of heads, one head for generating a vector by applying a weight to one of the embeddings.

17. The method of claim 15, wherein the embeddings of the tokens in the structured translated text are a second query input, a second key input, and a second value input to the first multi-head attention network.

18. The method of claim 15, wherein the output of the first multi-head attention network corresponds to a third query input received by the second multi-head attention network, and the encodings from the encoder correspond to a third key input and a third value input received by the second multi-head attention network.